The Next Tick on Nasdaq Stockholm: Predicting Price Direction in Limit Order Books Using Order Imbalance

Sammanfattning: We explore complete Level II limit order books for eight stocks listed on Nasdaq Stockholm during 2016 and investigate the use of the imbalance between bid and ask volumes in predicting the direction of price change in an ultra-high-frequency environment. Specifically, we test whether a top-of-the-book (Level I) measure of order imbalance and a deeper-in-the-book (Level II) measure can predict the direction of a change in the mid-price of a security for up to three events before the change occurs. Logistic models are used to fit the data and prediction power is judged based on the percent of correctly predicted observations out-of-sample. We find that order imbalance has statistically significant explanatory power in predicting price-change direction and that the logistic predictor considerably outperforms a naive one. In addition, we find that Level I order imbalance is more informative than Level II imbalance and that prediction power quickly decays in the order book. Finally, we compare our findings to those in Gould and Bonart (2016), on which the methodology in this paper is based. We find that unlike the case for US stocks, relative tick size does not play a role in explaining the shape of the order book for Swedish stocks. Rather, we argue that liquidity is the main factor driving order book behavior and price-direction predictability.